| from __future__ import annotations | |
| import json | |
| from scripts.aggregate_alpha_grid import BLOCKS, METHODS, SEEDS, aggregate | |
| def test_aggregate_reports_cells_and_paired_alpha_results(tmp_path): | |
| for alpha in (0.25, 0.5): | |
| for block_index, (model, task) in enumerate(BLOCKS): | |
| block_dir = tmp_path / f"alpha{alpha:g}" / "comparison" / f"{model}_{task}" | |
| block_dir.mkdir(parents=True) | |
| rows = {} | |
| for method_index, method in enumerate(METHODS): | |
| for seed_index, seed in enumerate(SEEDS): | |
| accuracy = 0.7 + block_index * 0.001 + method_index * 0.01 + seed_index * 0.0001 | |
| rows[f"{method}__rank32__seed{seed}"] = { | |
| "status": "complete", | |
| "eval_accuracy": accuracy, | |
| } | |
| (block_dir / "results.json").write_text(json.dumps(rows)) | |
| result = aggregate(tmp_path, alphas=(0.25, 0.5)) | |
| assert result["cell_counts"] == {"expected": 280, "complete": 280, "errors": 0} | |
| assert result["per_alpha"]["0.25"]["blocks"]["qwen3_xnli"]["fpeft_low"]["n"] == 5 | |
| comparison = result["per_alpha"]["0.25"]["paired"]["fpeft_low__vs__random_orthogonal"] | |
| assert comparison["n"] == 35 | |
| assert comparison["mean"] < 0 | |
| assert len(comparison["block_bootstrap_95"]) == 2 | |
Xet Storage Details
- Size:
- 1.35 kB
- Xet hash:
- fb31757d77a2494139c4f4dce8fced76b7745eed00b252d873f8fc6613bcb650
·
Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.